Likelihood of Bacterial Infection in Immunocompromised Patients Treated With IV Antibiotics for Possible Sepsis.

OBJECTIVES

Immunocompromised patients with possible sepsis are typically treated immediately with empiric antibiotics given their high risk for poor outcomes. Clinical presentations of sepsis, however, are protean and many nonbacterial infections and noninfectious conditions can present similarly. We assessed the likelihood of bacterial infections in retrospect in immunocompromised vs. nonimmunocompromised patients treated empirically for sepsis.

DESIGN, SETTING AND PATIENTS: We identified all adults treated for possible sepsis (blood culture draw, lactate measurement, and IV antibiotic administration) within 6 hours of emergency department (ED) arrival at nine hospitals between 2022 and 2024 and randomly selected 350 hospitalizations for structured medical record review, oversampling for immunocompromised status using administrative codes. Reviews were used to confirm immunocompromised status per Centers for Disease Control and Prevention criteria and determine post hoc likelihoods of bacterial infection.

INTERVENTIONS: None.

MEASUREMENTS AND MAIN

RESULTS

Of the 350 patients, 186 (53.1%) were confirmed as immunocompromised and 164 (46.9%) as nonimmunocompromised. Immunocompromised patients more commonly presented with fevers, chills or rigors (66.0% vs. 32.0%, p < 0.001) and ED clinicians more often documented concern for sepsis (74% vs. 62%, p = 0.01) compared with nonimmunocompromised patients. Definite or probable bacterial infection was identified in 92 of 186 (49.5%) immunocompromised patients vs. 97 of 164 (59.1%) nonimmunocompromised patients (p = 0.07). Among immunocompromised patients without definite or probable bacterial infection, common alternative etiologies included viral or fungal infection, febrile neutropenia without clinical evidence of bacterial infection, progression of malignancy, and medication toxicity, whereas chronic heart or lung disease exacerbations, hypovolemia, and hypervolemia predominated among nonimmunocompromised patients.

CONCLUSIONS

Only about one-half of immunocompromised patients treated empirically for possible sepsis had definite or probable bacterial infection in retrospect, compared with approximately three-fifths in nonimmunocompromised patients. These findings underscore the complexity of sepsis diagnosis, particularly in immunocompromised patients, and highlight important differences in the conditions that mimic bacterial sepsis in immunocompromised vs. nonimmunocompromised hosts.